Who’s calling at this hour ? Local Sideral Time and telephone telepathy

Who’s calling at this hour ? Local Sideral Time and telephone telepathy

Le biologiste Rupert Sheldrake a développé différentes expériences afin de mettre en évidence des perceptions psi au sein de situations quotidiennes. C’est dans cette perspective qu’il a étudié un phénomène rapporté par de nombreuses personnes, à savoir l’impression de pouvoir de déterminer qui les appelle avant de décrocher leur téléphone. S’agit-il d’une impression subjective reposant sur le hasard ou d’une réelle précognition ?
Par ailleurs, un autre parapsychologue, James Spottiswoode, a proposé une hypothèse provenant d’analyses d’expériences antérieures : le Local Sideral Time (LST ; l’heure calculée en un endroit de la Terre en fonction de sa position par rapport aux étoiles « fixes », et non par rapport au soleil) serait un paramètre à prendre en compte pour améliorer les perceptions psi.
La recherche de Dick Bierman et Eva Lobach, proposée ci-dessous, est à la rencontre de ces deux approches : elle a pour objectif d’étudier expérimentalement la télépathie (« par » téléphone) et l’influence du LST sur cette dernière.
Nous attirons l’attention des lecteurs sur le fait qu’on ne peut conclure d’une seule recherche de ce type que la télépathie existe et quelle est l’importance du LST. L’intérêt de cette recherche, publiée dans les proceedings du congrès 2004 de la Parapsychological Association, réside dans le fait qu’elle illustre une tentative – parmi bien d’autres – de mise en évidence de paramètres favorisant l’émergence et l’étude des perceptions psi.


Can we guess who is calling us on the phone before picking up, and does local sidereal time (LST) affect how
often we guess right? Reviews of anomalous cognition studies have shown that effect sizes are highest around
13.30 LST (Spottiswoode, 1997). A post-hoc analysis of telephone telepathy data of Sheldrake (2003) also showed
a peak at that time. LST (peak or non-peak) was an independent variable in our prospective telephone telepathy
study. Six women who indicated they often experienced telephone telepathy were selected to participate. Each
participant chose four close friends or relatives to act as callers. All completed a total of 36 trials; six sessions of
six trials each, three sessions at peak time (between 8.00 and 9.00 local time) and three at non-peak time (between
17.30 and 18.30 local time). One of the experimenters was at the participant’s home during the sessions. The
experimenter made sure no irregular communication was going on and logged times of the calls and responses of
the participant. At a different location another experimenter used a dice to select a caller about five minutes
before the scheduled trial. Then he or she contacted the caller who was asked to call the participant in five
minutes and to concentrate his or her thoughts on the participant for the last two minutes before the call was
made. When the phone rang at the participant’s home, the participant guessed who she thought was calling
before picking up. Analyses show a significant over-all scoring rate of 29.4% (p = .05). Almost all of this effect
originates from the sessions at peak time with a scoring rate of 34.6%. Exploratory analyses show that a stronger
emotional bond between particpant and caller is associated with a higher hitrate. It is concluded that results
provide tentative support for the hypothesis that Local Sidereal Time is related to a phenomenon like telephone
telepathy. In addition, the results are in support of the existence of telephone telepathy. Other explanations of
the anomalous effect cannot be ruled out, such as precognition, retro psychokinesis by the experimenter or the
participant so the dice throw would coincide with the particular caller the participant would guess, or
clairvoyance of the dice throws. Future studies should aim at teasing apart the supposed effects of LST and local
time on ‘telephone telepathy.’


You are at home. The phone rings. And you know who is calling before answering. This kind of
‘telephone telepathy’ appears to be an ordinary phenomenon, although some of us appear to be better at it
than others. A number of studies with pre-selected individuals who performed above chance in a pilot study
showed evidence that some people are quite good at guessing who is calling them, especially when the callers
are good friends or relatives (Sheldrake, 2003).
Interestingly, there is evidence that telephone telepathy, which is considered to be a form of anomalous
cognition (AC), not only varies among individuals, but also varies along the hours of the day, or more
accurately, by local sidereal time (LST). In a large sample of anomalous cognition studies, Spottiswoode
(1997) found that effect sizes were highest around 13.30 h LST. Although there is no obvious explanation
for such a finding, the same distribution of effect sizes was confirmed in a separate sample of anomolous
cognition studies. Submitting Sheldrake’s telephone telepathy data to the same procedure, these again
showed the largest effect sizes around 13.30 h LST (Spottiswoode, 2003).
These findings are very suggestive, but they are post-hoc analyses, and therefore a prospective study is
much needed. Our goal was to replicate the telephone telepathy study of Sheldrake, and use LST as a
within-subjects independent variable.
Although we implemented a few changes in order to obviate criticisms aimed at Sheldrake’s (2003)
studies, the general lay-out of our study was very similar. Participants were at home, since a familiar environment is assumed to foster telepathic sensitivity, and would receive several phone calls on their
landline phone (without number display), in a previously arranged period of time. The participant knew
that the dice thrown by the experimenter at some different location would determine which one of four
different callers would be calling. Before picking up, the participant would state clearly who it was.
Instead of audio- or videotaping the sessions as Sheldrake did, we chose to have one of the other
experimenters visit the participant, in order to check if there were no other types of communication going
on before the phone calls were being made so as to preclude any suspicions of hidden accomplices.
All participants took part in six sessions, three at peak LST (peak condition), around 13.30 h LST, and
three at non-peak LST (non-peak condition) around 00.00 h LST. Participants were not informed of the
LST hypothesis.
Although Spottiswoode’s analyses showed that the lowest AC effect sizes occurred around 18.00 h LST,
we chose 00.00 h LST as our non-peak time for a pragmatic reason. When this experiment was conducted
(December 2003 / January 2004, Amsterdam, The Netherlands), peak LST was between 8.00 and 9.00 AM,
and Spottiswoode’s low at 18.00 LST would be around lunch time. We reasoned that it would be difficult to
organize sessions at the participants’ homes around that time, and instead chose to conduct the non-peak
sessions around 00.00 LST, which was between 6.30 and 7.30 PM local time. According to Spottiswoode’s
analyses, this would be a time of about average effect sizes.



Participants were recruited through e-mails sent to friends and acquaintances of the experimenters. They
had to meet four requirements: they had to be available at the scheduled times, live close enough to one of
the experimenters, find four good friends or relatives to cooperate, and have experienced that they often
guessed correctly who was calling them on the phone.
Six women, aged 16, 19, 20, 21, 21, and 54 were selected. They received €40 for taking part in all six
sessions. The callers, four for each participant, 24 in total, each received €15. Participants were at home
during all sessions. Callers could be anywhere, as long as they could be reached by phone.


Local Sidereal Time. Local sidereal times were calculated by means of on online scheduler, Local Sidereal
Time – Monthly experiment scheduler for PSI research (Melssen, 2003).
Number of sessions and trials. Each participant took part in six sessions, three at peak and three at non-peak
LST. Each session consisted of six trials, so there were about ten minutes in between the calls. Participants
were each called 36 times, the total number of trials was 216.
Emotional bond. Before the start of the first session, the participants indicated their emotional bond with
each of the four selected callers on a five point scale from 1, ‘good emotional bond’, to 5, ‘very strong
emotional bond’.

Logs. There were two logs, one was kept by the experimenter visiting the participant (‘visiting
experimenter’), and the other one was kept by the experimenter selecting the callers (‘control experimenter’).
Both logs bore the name of the participant and names of the callers, date, LST condition and session
The control experimenter wrote down the time at which each call was made to the selected caller, and
who was selected to make the call. Any irregularities, such as when a caller couldn’t be reached, were also
The visiting experimenter wrote down the time the participant received the call, the response of the
participant and whether it was a hit or a miss, according to the participant.

Randomization. A number between one and four was pre-assigned to each caller. A dice thrown by the
control experimenter determined which caller had to make a particular call. Five and six were disregarded.


Just before the scheduled time, the visiting experimenter arrived at the participant’s home and checked if
everything was in order. He or she then retreated to a different room, within hearing distance of the
participant, but out of sight, so as to disturb the home environment of the participant as little as possible.
The callers weren’t at any specific location, but were asked to be available by phone at the scheduled

At the start of each session, the control experimenter contacted the four callers and the participant to tell
them that the session had started and to check if everything was ready at the participant’s home. If it turned
out that not all callers were available at the start of the session, the procedure was still followed through.
Before each trial, the control experimenter threw a dice to select the caller for that trial. The caller was
told to call the participant in five minutes, and to concentrate his or her thoughts on the participant for the
last two minutes before making the call. If the caller’s line was busy or if the caller didn’t answer, the control
experimenter contacted the caller again in three minutes and asked the caller to call the participant in two
minutes. If the line was still busy or the caller was still unavailable, the experimenter would throw the dice
again to select another caller.
When the phone rang at the participant’s home, the participant said who she thought was calling, and
the response was logged by the visiting experimenter. The participant then answered the phone and told the
visiting experimenter whether she was right or wrong.
After six trials the control experimenter contacted the callers to tell them that the session was over, to
thank them for their cooperation and to remind them of the next session.


Session integrity

Logs of the two experimenters were put together in one datafile. In two cases we found a discrepancy
between the two logs. According to the logs, the participant’s guess and the caller were the same, so it
should have been a hit, but the visiting experimenter had logged it as a miss. These two trials were further
considered as missing values.
The logs showed that in ten of the 36 sessions, one or more of the callers were partly or totally
unavailable for various reasons. They could have missed the appointment entirely, or could have forgotten
to turn there mobile on in time. In a number of cases, a caller was not available during part of a session, e.g.,
because she was taking a shower. It seemed unlikely, though not impossible, that the participant would
know about the absence of one of the callers. The chances of guessing correctly if one is aware of the
absence of a caller is 1/3 rather than the a priori hit probability of 1/4. Therefore we decided to analyse the
data twice: once with all sessions included (Table 1), and once again with only the regular sessions (Table 2).

Distribution and randomness

To acquire an indication of the ‘randomness’ of the dice throws and the participants’ guesses, we
checked the frequency distributions for all trials and all consecutive pairs of trials.
The frequency distribution of dice throws and of the participants’ guesses did not differ from the
expected 25% for each of the four options, χ2 (3) = 3.53, p > .3 and χ2 (3) = 4.51, p > .2 respectively.
As an indication of the randomness of the sequences, we looked at all pairs of consecutive dice throws
and guesses. If the sequences are random, consecutive pairs of similar numbers (in this case 11, 22, 33, and 44) are expected to make up 25% of the total of all consecutive pairs, while 75% of the pairs would be

For the dice, 35 (20%) of the 179 consecutive pairs contained similar numbers, less than expected
according to chance. The difference with chance expectation was not statistically significant, however, χ2 (1)
= 2.83, p =.09.

For the guesses, 52 (29%) of the 179 consecutive pairs contained similar numbers, more than expected
according to chance, but again the difference was not statistically significant, χ2 (1) = 1.47, p = .23.
Although these tests cannot firmly establish whether the sequences are truly random, it should be noted
that the deviations from chance expectation for dice throws and guesses are in opposite directions. This
means that it is highly unlikely that peculiarities of the sequences per se could account for an above chance

Over-all hitrates

tableau 1

Table 1: Number and expected number of hits at peak LST sessions, non-peak LST sessions and total

Notes: Total number of sessions was 36, 18 at peak and 18 at non-peak LST. Each participant took part in 6 sessions (3 at peak and 3 at non-peak LST), and received 6 calls per session. For two trials, one peak-trial for participant 4 and one non-peak trial
for participant 5, the logs of the experimenters showed discrepancies. These trials are therefore registered as missing values.
Hexp indicates the expected number of hits.
Pooling peak and non-peak conditions together, the over-all scoring rate is 29.4% with all sessions
included (Table 1) and 32.7% for the subset of sessions that confirmed strictly to the protocol (Table 2).
Testing individual scoring rates of the participants against expected scoring rates (25%), a paired-samples ttest
shows that scoring rates are significantly above chance, t(5)=2.01, p=0.05 (one-tailed) when all sessions
are included (Table 1), and also for the subset of the 26 regular sessions (Table 2), t(5)=5.48, p<.005 (onetailed). Tableau 2

Table 2: Number and expected number of hits at peak LST sessions, non-peak LST sessions and total for the 26 sessions where callers were available throughout

Note: Results are based on 15 peak sessions and 11 non-peak sessions, a total of 26 sessions. See also the notes at Table 1.

Telephone telepathy and LST

As Table 1 shows, the scoring rate was higher than expected in the peak condition (34.6%), while it was
around chance (25.2%) in the non-peak condition. To test the statistical significance of this 9.4% difference,
we compared scoring rates in peak and non-peak condition for each participant. A paired-samples t-test
showed that scoring rates were marginally significantly different between peak and non-peak condition,
t(5)=1.60, p=.09 (one-tailed).
For the sessions in which all callers were available throughout (Table 2), the average scoring rate was
35.6% for the peak and 28.8% for the non-peak condition. This time, we could only compare five
participants, because all non-peak sessions of one participant had to be dropped from analysis. The
difference of 6.8% between peak and non-peak condition was not statistically significant.

Emotional Bond and Hitrates

The average emotional bond with the callers was M=3.71 (sd=1.12), indicating that the participants had
indeed a strong emotional bond with the callers. To explore the extent of response bias due to emotional
bond, we calculated the correlation between emotional bond and the number of times a participant guessed
a particular caller. This correlation was performed using all 214 trials, after normalizing the variables, and
turned out to be not significantly different from zero, r=.20, ns.
To explore if emotional bond between participant and caller was related to success at guessing, we again
first normalized the emotional bond ratings and scoring rates for each participant separately, including all
214 trials. The correlations between emotional bond and scoring rate ranged from -0.73 to 1.00. Five of the
six participants showed a positive correlation. Overall correlation between emotional bond and scoring rate
was r=.41, p<.05. Controlling for response bias (the number of times the participant guessed a particular caller), the correlation between emotional bond and hitrate still reached about the same value, r=.39, p=.07.


Results show weak support for the hypothesis that Local Sidereal Time affects an anomalous cognition
phenomenon like telephone telepathy. Our sample was rather small, and the large number of sessions that
showed irregularities further weakened the power of our statistical analyses. However, the results are still
promising and replications are warranted.

Future studies should take into account that this single study could not separate the effects of local time
and local sidereal time. It is possible that the tentative support for the LST hypothesis must be attributed to
the fact that the LST peak sessions were all conducted early in the morning. Although we are not aware of
any evidence that anomalous cognition effects are related to the time of day, it cannot be excluded. When
the LST peak times are during the evening, however, it will always be summer at this location of the globe.
Type of season is another factor that could potentially affect telephone telepathy. Replications in other
continents are therefore especially helpful to gather more evidence concerning the effect of LST on
anomalous cognition.

Results are more supportive of the existence of telephone telepathy. Overall, our participants guessed
correctly more often than would have been expected according to chance. The evidence is not as impressive
as the 45% results reported by Sheldrake (2003), however, which may have been due to his preselection of
participants through a pilot study, while we employed a rather loose selection criterion by including only
those participants who reported to have had experiences of telephone telepathy in the past.

The correlation between emotional bond and hitrate in our study showed that even in a relatively
homogeneous group of friends and relatives, people still appear to be better at guessing those callers with
whom they have the strongest emotional bond. This result is in line with Sheldrake’s findings.

The effect size reported here is very close to the typical effect size found in Ganzfeld experiments. Thus
the required number of trials from a power perspective (power=0.5) is between 200 and 250 to obtain a result significant at the 5% level. However, there is a huge difference in time investment to obtain this
result. We ran 12 trials per day while in Ganzfeld research it is typically 1.
Could the telephone telepathy in this experiment be explained by more ‘ordinary’ circumstances during
the experiment?
There are some concerns.
Logs were kept by hand and experimenters could have made errors. In fact, we identified two errors (in a
total of 214 trials), where the response of the participant and the name of the caller matched, but the log of
the visiting experimenter nevertheless registered a miss. We treated these two instances as missing trials.
Although there may have been one or two additional but undiscovered mistakes in registering the name of
the caller or the actual response of the participant, it seems highly unlikely that a hit or a miss was logged
inaccurately. There were only six trials in an hour, and both the participant and the experimenter were very
keen on the outcome of every trial.
Could the callers have notified the participant before the call was made? This seems possible, using
mobile phones in the buzz mode. A potential caller could have a code of 1, 2 or 3 buzzes before hanging up.
The visiting experimenter was aware of this option, however. Furthermore, there was not much time left for
the caller between being contacted by the control experimenter and making the actual call to the
The participant could know some callers to be late, and this might produce subtle timing differences
between callers. We had discussed the adoption of a random timing procedure, so the control experimenter
would contact the callers at varying time intervals, instead of keeping to a ten-minute interval, as was the
case now. However, we suspected that the timing of trials would be subject to some fluctuations anyway, e.g.,
when it would take the experimenter longer to throw a number under five, or when the caller was not
directly available. Nevertheless, timing differences between callers might still provide an explanation of
correct guesses at some trials, and future studies should consider a random timing procedure, in
combination with a less cumbersome randomization procedure.
Lastly, there might have been a collusion between student experimenters and student subjects. Only
replications with extra technical security measures might be able to make this type of explanations less

It seems, nonetheless, that the results of this study provide further evidence for the existence of some
anomalous effect, but should we call it ‘telepathy’? In spite of using instructions for the caller to concentrate
and think about the person to be called, instructions consistent with the telepathy model, some other
options are possible.

First, there could be precognition at work. Precognition appears to be dependent on the time between
the guess and the feedback and in this study, this time was short and hence could have produced good
precognition results. Second, psi could have entered at the experimenter randomization level. That is, the
dice throwing was such that it correlated with what the participant would say anyway. This could be seen as
retro psychokinesis by the experimenter or the participant. And finally, it could be clairvoyance: The
participant could be ‘tuned’ to the dice being thrown by the experimenter and know who was selected for
the call as soon as ‘the die was cast.’ Considering this option, we inspected the trials in which the dice
indicated a particular caller who later turned out to be unavailable. This happened at eight trials. In five of
these eight trials, the participant guessed the name of the caller who should have called if he had been
available (exact binomial p=.03). Of course, instead of clairvoyance, the ‘missed hits’ could also indicate
telepathy with the control experimenter. Note that the finding that participants still mentioned the names
of callers who turned out to be unavailable, counters the suggestion that ‘psi’ effects could have resulted
because the participant was somehow informed of the absence of one of the callers.

In light of the generally unsuccessful replications of psi research at the university of Amsterdam the
current results should be considered as a recommendation for further exploration of the telephone telepathy


Melssen, F. (2003). Local Sidereal Time . Monthly experiment scheduler for PSI research.

Sheldrake, R., & Smart, P. (2003). Videotaped experiments on telephone telepathy. Journal of Parapsychology, 67, 187-206.

Spottiswoode, S.J.P. (1997). Apparent association between effect size in free response anomalous cognition experiments and local sidereal time. Journal of Scientific Exploration, 11, 1-17.

Spottiswoode, S.J.P. (2003) .Telephone telepathy.. Analysis of data for LST and GMF effects. Private communication.

Adresse pour correspondance : Dick Bierman,
University of Amsterdam, Roetersstraat 15,
1018 WB Amsterdam, The Netherlands.
E-mail: d.j.bierman@uva.nl